An Intelligent Rule-Based Compensator for Control System Input Nonlinearities

The paper presents an Al alternative to classical control compensation procedures for input nonlinearities. Specifically, a rule-based system is developed which compensates for dead-zone phenomena in a position control servo-mechanism. The rule-based system compensator is based around a supervisory module which is designed to detect and avoid undesirable nonlinear control phenomena such as steady state errors, limit cycling or excursions into unwanted zones. The heuristics within the supervisor also allow abnormal situations to be detected and avoided. The technique is thus potentially far more powerful than conventional non-linear compensation procedures such as `dither'. As such, it constitutes an intelligent compensator for system input nonlinearities which may be generalizated to other nonlinearity types.